Anthropic Unveils Sonnet 4.5: The New King of Coding AI
In May, Anthropic introduced its AI systems, Opus 4 and Sonnet 4, showcasing remarkable advancements. Now, just six months later, the company has launched Sonnet 4.5, claiming it to be the best coding model in the world.
Significant Performance Improvements
Anthropic’s Sonnet 4.5 surpasses earlier models and competition, including Google’s Gemini 2.5 Pro and OpenAI’s GPT-5. In the OSWorld benchmark suite, Sonnet 4.5 achieved an impressive score of 61.4%, outperforming Opus 4.1 by a staggering 17 percentage points.
Enhanced Longevity for Multi-Step Projects
One of the standout features of Sonnet 4.5 is its ability to autonomously tackle multi-step projects for over 30 hours. This is a significant enhancement compared to the roughly 7 hours that Opus 4 provided, marking a vital milestone in agentic AI development.
Unmatched Safety Measures
Safety has been a primary focus for Anthropic with Sonnet 4.5. The company claims that this model is its safest to date, having undergone extensive safety training to minimize issues like “sycophancy, deception, and delusional thinking.” Additionally, enhanced protections against prompt injection attacks ensure a more reliable user experience.
Quality-of-Life Upgrades
Alongside Sonnet 4.5, Anthropic is rolling out improvements across its Claude product stack. Users will benefit from a revamped terminal interface in Claude Code, which now includes a checkpoints feature to save and revert progress easily. This update aims to enhance user interactions and streamline coding tasks.
Affordable API Pricing
For developers interested in using Sonnet 4.5, the API pricing remains unchanged at $3 per million input tokens and $15 for the same amount of output tokens. This pricing structure continues to make Anthropic’s offerings competitive in the AI market.
Conclusion
The release of Sonnet 4.5 highlights Anthropic’s commitment to advancing AI capabilities while prioritizing safety and user experience. With its record-breaking performance and integrated safety measures, Sonnet 4.5 sets a new standard in the realm of coding models.
Related Keywords: AI models, coding AI, Anthropic Sonnet, safety in AI, AI benchmarks, intelligent agents, software development tools.
